How to Check If ChatGPT Recommends Your Business: A 10-Minute Test
Ask the same three questions three times on ChatGPT and Google's AI Mode, count how often you get named, and repeat monthly. Here is the exact test.
Key Takeaways
- 45% of US consumers used AI tools for local business recommendations in the past year, up from 6% the year before (BrightLocal, 1,002 US adults).
- ChatGPT was the most-used tool for those recommendations, at 31% of consumers (BrightLocal).
- ChatGPT recommended 1.2% of locations, Gemini 11.0% and Perplexity 7.4%, against 35.9% for Google's local 3-Pack (SOCi, 2,751 brands).
- In SparkToro's 2,961 test runs, there was under a 1 in 100 chance that ChatGPT or Google's AI would give the same list twice.
- The test below takes 18 runs and about 10 minutes, and ends in one number: your named rate.
Most owners have never typed their own category and city into ChatGPT and watched what comes back. A roofer in Tulsa, a motel owner near an airport and a dentist in a strip mall all have the same blind spot: customers are asking an AI who to call, and nobody on the business side has seen the answer.
How do I check if ChatGPT recommends my business?
Ask ChatGPT and Google's AI Mode the questions a customer would ask. Run each question three times in a fresh chat. Count how many answers name your business, then divide by the total runs. That percentage is your named rate. It takes about ten minutes and costs nothing.
Want a quick read before you run the full test? The free AI Visibility Site Scan scores your site for AI search in about a minute, with no email needed for the score. Use it alongside the prompt test, since it cannot show what ChatGPT actually says about you.
An AI visibility check is a test of whether, and how, AI assistants name your business when a customer asks for a recommendation. The result is a frequency (how often you appear), not a ranking.
Why three runs and not one? Because AI answers do not repeat. SparkToro and Gumshoe.ai had 600 volunteers run 12 prompts through ChatGPT, Claude and Google's AI, for 2,961 runs in total. The chance of getting the same list twice was under 1 in 100, and the chance of the same order was about 1 in 1,000. A single check can flatter you or insult you, and neither means anything.
Which prompts should I test?
Test three prompts that match how a real customer asks: one broad, one urgent, one with a constraint. Use your actual service, city and the words customers use. Skip prompts with your business name in them for now. Those only test whether AI knows you exist.
- Broad: "Who is the best roofing company in Tulsa?" or "Best dentist in Scottsdale for a nervous patient?"
- Urgent: "HVAC repair near me, open on a Sunday" or "Burst pipe, who can come tonight in Dayton?"
- Constraint: "Family-run motel near the airport with pet-friendly rooms" or "Roofer in Tulsa who handles insurance claims"
Write the prompts down and reuse them word for word every month. Changing the wording changes the answer, which wrecks the comparison. The 600 volunteers in the SparkToro study wrote very different prompts for the same intent (the authors measured an average semantic similarity of 0.081 across 142 headphone prompts), so a fixed script is how you keep your own test consistent.
How do I run and score the test?
Open a fresh chat for every run, so earlier answers do not carry over. Log out of the tool or use a private window if you can. Your own history may tilt results toward businesses you have already looked at, and we have not measured by how much. Then work the grid below.
- Run each of the three prompts three times on ChatGPT. That is 9 runs.
- Run the same three prompts three times on Google's AI Mode. That is 9 more, 18 in total.
- For each run, write one line: named or not named, who was named instead, and whether the details about you (phone, hours, service area) were right.
- Named rate = runs where you appeared, divided by 18. Write that number down with today's date.
- Copy the competitors who show up most often. They are the businesses AI currently trusts in your market.
Ignore position. If you were named first once and third once, that is one data point twice, not a trend. SparkToro's authors are blunt about it: any tool that gives a "ranking position in AI" is, in their words, full of baloney. They describe visibility percentage across many prompts, run many times, as the reasonable metric. We use 18 runs because it fits ten minutes. That count is our rule of thumb, not a statistical standard. SparkToro lists how many runs are enough as an open question.
How reliable is an 18-run test?
It is reliable enough to tell you whether you are in the conversation at all, and not precise enough to detect small month-to-month moves. A named rate of zero across 18 runs is a clear signal. A move from 22% to 28% is noise. Repeat the test monthly and watch the direction over a quarter.
Read the published benchmarks with the same care. The SparkToro study was run by people who say they are not credentialed researchers, was not peer reviewed, collected data over about seven days in November and December 2025, and one author is tied to an AI tracking startup. SOCi's 2026 Local Visibility Index is vendor-run and covers multi-location brands: 2,751 brands and roughly 350,000 locations. It found ChatGPT recommended 1.2% of locations, Gemini 11.0% and Perplexity 7.4%, against 35.9% appearing in Google's local 3-Pack. Those are context for your number, not targets for it.
Should I test by hand, use a scan, or pay for a tracker?
Start by hand, because it shows you the actual answers and the competitors in them. Add a scan to check why your site may be hard for AI to read. Pay for a tracker only once you are repeating the test often enough that your time costs more than the subscription. The three options answer different questions.
| Manual prompt test | Automated site scan | Paid tracking tool | |
|---|---|---|---|
| Time | About 10 minutes per round | About 60 seconds | Setup once, then automatic |
| Cost | Free | Free | Monthly subscription (check each vendor's pricing page) |
| What it covers | The real answers for your own prompts | Whether your site gives AI clear facts to read | Many prompts, many tools, on a schedule |
| What it misses | Small samples and your own wording | What AI actually says about you today | Local nuance, unless you set the prompts yourself |
| Best for | Your first read and every monthly check | Finding site-side fixes | Owners or agencies tracking many locations |
Why does this matter now?
Because customers moved first. BrightLocal's 2026 Local Consumer Review Survey, a panel of 1,002 US adults published in March 2026, found 45% of consumers used AI tools for local business recommendations in the past year, up from 6% the year before. ChatGPT was the leader at 31%. The survey covers all business categories, so it describes consumers in general, not your trade.
The next step is already shipping. On October 6, 2026, ServiceTitan announced Homh, a platform that connects homeowners and their AI agents with selected ServiceTitan contractors, and says it works with ChatGPT, Gemini and Claude. Only participating contractors are listed. If you are not one of them, the prompt test is how you see what homeowners are being told instead. Homh covers ServiceTitan contractors only.
What if my named rate is zero?
Zero is common and fixable. It usually means the AI could not confirm who you are from consistent facts on the web. Start with your name, address, phone and hours matching everywhere, then your rating and your website copy. We lay out the order in why ChatGPT doesn't recommend your business and collect the supporting numbers in AI search statistics for local businesses.
If you run a lodging business, the same test applies with different prompts. See how independent hotels get found in AI search and the hotel AI search FAQ. Our free AI Visibility Site Scan checks the site-side basics in about a minute, and our AEO service covers the full program.
Frequently Asked Questions
How do I know if ChatGPT recommends my business?
Ask it the questions a customer would ask, in a fresh chat, and repeat each question three times. Count the runs where your business is named and divide by the total. Because answers vary, a single check tells you very little. SparkToro found under a 1 in 100 chance of the same list appearing twice.
How many times should I run each prompt?
We use three runs per prompt across three prompts and two tools, which is 18 runs and about ten minutes. That is a practical rule of thumb, not a statistical standard. SparkToro's authors list the number of runs needed for sound results as an open question.
Does my position in the answer matter?
Not much. Order changes from run to run, with roughly a 1 in 1,000 chance that two runs return the same order in SparkToro's data. Track how often you appear, not where. A named rate is steadier than a rank.
Do I need a paid tool to track AI visibility?
No. A manual test costs nothing and shows you the real answers and the competitors named in them. A paid tracker saves time once you repeat the test often or manage many locations, but it adds nothing to your first read.
How often should I rerun the test?
Monthly, with the same prompts word for word. Log the date and the named rate each time. Small changes between months are noise, so judge the direction over a quarter before changing your plan.